The present invention generally relates to charged particle beam exposure methods and apparatuses, and more particularly to a charged particle beam exposure method which is suited for realizing a high resolution electron beam exposure at a high throughput, and to a charged particle beam exposure apparatus which employs such a charged particle beam exposure method.
Recently, there is much attention on exposure techniques using a charged particle beam such as an electron beam because of the need to accurately expose extremely fine patterns to meet the increasing integration density of semiconductor integrated circuit devices. In order to realize a high throughput, the charged particle beam is shaped into a beam having an arbitrary cross sectional shape and the shaped beam is irradiated on a wafer. It is desirable that the numerical aperture is large in order to realize a high resolution, and it is desirable that the charged particle beam travels a short distance to the wafer so as to reduce the Coulomb interactions of the charged particle beam. For these reasons, an optical system of the charged particle beam has a short focal distance.
An electron beam exposure apparatus normally uses an electron beam which is shaped to have a rectangular cross section or a cross section having an arbitrary shape, and the shaped electron beam draws the pattern on the wafer. FIG. 1 shows the column structure of an example of a conventional electron beam exposure apparatus.
FIG. 1 shows a cross sectional view of the column structure in a vicinity of the wafer. An electron beam EB travels from the top to bottom along an optical axis O, and is irradiated on a wafer W. An electromagnetic lens is provided to converge the electron beam EB on the wafer surface. The electromagnetic lens includes an electromagnetic lens coil LC which is coupled to an iron yoke IY, and pole pieces PP are coupled to the tip end of the iron yoke IY.
In addition, electromagnetic deflection coils EC and electrostatic deflection electrodes EE are provided to deflect the electron beam EB. In order to shorten the focal distance, the electromagnetic deflection coils EC and the electrostatic deflection electrodes EE are arranged on the inner side of the electromagnetic lens.
In the example shown in FIG. 1, a support part S is arranged on the inner side of the electromagnetic lens, and the electromagnetic deflection coils EC are mounted on this support part S. Further, the electrostatic deflection electrodes EE are arranged on the inner side of the support part S.
The electron beam EB is deflected by the electromagnetic deflection coils EC when deflecting the electron beam EB for a relatively large amount on the order of several mm, for example. On the other hand, the electron beam EB is deflected by the electrostatic deflection electrodes EE when deflecting the electron beam EB at a high speed for a relatively small amount on the order of 100 .mu.m, for example.
In the case of the electron beam exposure which exposes a relatively large area in one exposure, the Coulomb interactions of the charged particles cause problems. The focal distance of the optical system is shortened in order to eliminate the limit of the resolution caused by the Coulomb interactions. However, the deflection efficiency of the deflector deteriorates if the focal distance is shortened. Accordingly, in the electron beam exposure system having the shortened focal distance, a large current must be applied to the electromagnetic deflection coils EC in order to obtain a desired amount of deflection.
FIG. 2 shows a perspective view of the electromagnetic deflection coils EC. The electromagnetic deflection coils EC have a saddle shape, and are fixed at mutually confronting positions on the outer periphery of the support part S. The support part S has a cylindrical shape and is made of ceramics or the like.
According to the conventional electron beam exposure apparatus, the electron beam EB is deflected using the electromagnetic deflection coils EC provided in a plurality of stages, and the pattern is drawn on the wafer W by scanning the wafer surface by the deflected electron beam EB. The electromagnetic deflection coils EC are divided into two systems, that is, X and Y systems, depending on the operating direction. The electromagnetic deflection coils EC are coupled in series within each system. The electromagnetic deflection coils EC of the two systems receive driving currents from independent driving circuits.
For example, a current on the order of .+-.2 A is required to deflect the electron beam EB with a deflection efficiency of approximately 2.5 mm/1A in a main deflection region which is 2 mm.times.2 mm. Hence, if each electromagnetic deflection coil EC is formed from a copper wire having a diameter of 0.5 mm, the resistance thereof becomes approximately 1.5 .OMEGA..
In order to avoid the Coulomb interactions of the charged particle beam, it is necessary to shorten the focal distance of the optical system of the charged particle beam. But if the focal distance is shortened. the deflection efficiency deteriorates, and a larger current is required if the same amount of deflection is to be obtained with the shortened focal distance.
In addition, in order to operate the electromagnetic deflection coils EC at a high speed, it is necessary to reduce the inductance, and thus, it is desirable to reduce the area of the electromagnetic deflection coils EC.
On the other hand, heat is locally generated within the column structure if the charged particle beam is processed using the electromagnetic deflection coils EC having the above described arrangement, and such a generation of heat is unavoidable. The generated heat ranges from several W to several tens of W, for example.
When the charged particle beam exposure apparatus of the type described above is used, the deflection position of the charged particle beam and the focal position (or point) of the charged particle beam drift with the operating time of the charged particle beam exposure apparatus. It may be regarded that the following causes the drift of the deflection position and the focal position of the charged particle beam.
(a) Changes in the outputs of an amplifier and a lens power source;
(b) An eddy current flowing to peripheral metal parts due to a change in the magnetic flux generated by the deflection coils;
(c) Charge-up of parts through which the charged article beam passes;
(d) Changes in the position and dimension of the deflection coils with time; and
(e) Changes in the positions and dimensions of the deflection coils themselves, bobbins and other parts due to a temperature change caused by the heat generated from the deflection coils.
The present inventors initially doubted the possibility that the response characteristic of the deflection coil greatly deteriorates with the operating time of the charged particle beam exposure apparatus. However, the output of the amplifier had not changed, and the set time constant remained the same. In addition, the inductance of the deflection coil also remained approximately the same. In other words, even though the deflection magnetic field was set, some factor changed the beam position. But it seemed impossible for the eddy current to change before and after the exposure when the same deflection was made.
Furthermore, the beam position on the optical axis is reproducible for a long time if the charged particle beam is not deflected. For this reason, it seemed impossible for the charge-up to cause the deterioration of the response characteristic of the deflection coil.
The remaining possibility was the changes in the positions and dimensions of the deflection coils themselves, the bobbins and other parts such as the pole pieces due to temperature changes caused by the heat generated from the deflection coils. As described above, the heat of several W to several tens of W may be generated from the electromagnetic deflection coil, and the radiation effect is poor if the electromagnetic deflection coils are arranged in a vacuum surrounding.
In the converging deflection system having the shortened focal distance, the lens magnetic pole becomes small because of the need to make the lens magnetic field strength large and the deflection magnetic field strength large. Consequently, the deflection coils which are provided on the inside must be arranged without a gap within a space which is narrow in both the direction of the optical axis and the radial direction. As a result, the part which holds the deflection coils is made extremely thin, and the heat capacity thereof is reduced to several fractions of that of the conventional case.
In addition, the difference between the radii of the deflection coils in the X and Y systems is small. But although the size of the deflection coil is reduced, the thickness of the wire member is approximately the same as that of the conventional case. For this reason, the thickness of the deflection coil in the direction in which the wire is overlapped is large, and the inner turns of the wire of the deflection coil are covered by the outer turns of the wire.
For this reason, even if the outer side of the deflection coil were air-cooled, the air-cooling effect would greatly differ between the outer turns of the wire and the inner turns of the wire. The inner turns of the wire of the deflection coil virtually cannot be air-cooled directly, and the cooling is in effect made via the thermal conduction of the bobbin. Hence, the cooling effect of the inner turns of the wire of the coil and the cooling effect of the outer turns of the wire of the coil greatly differ.
If the deflection coil were made using a thin wire member in order to improve the air-cooling effect, the amount of heat generated from the deflection coil would increase. As a result, the deflection accuracy of the charged particle beam would further deteriorate due to the thermal expansion of the wire member itself and the thermal expansion of the bobbin on which the deflection coil is attached.
In a case where a copper wire member is used for the deflection coil and the bobbin is made of ceramics, positional errors of 0.34 .mu.m and 0.16 .mu.m may respectively be generated in a main deflection region of 2 mm.times.2 mm when the temperature rises by 10.degree. C., because the coefficient of thermal expansion of the deflection coil is 1.7.times.10 cm.sup.-5 and the coefficient of thermal expansion of the bobbin is 8.times.10 cm.sup.-6 in this case.
The deflection coil and the bobbin are actually attached to each other, and it may be anticipated that the amount of positional error will take a value between 0.34 .mu.m and 0.16 .mu.m. Even if this anticipated value of the positional error were 0.2 .mu.m, the positional error in the main deflection region would be 0.4 .mu.m at the maximum. Such a positional error is too large when forming patterns on the submicron order.
There is yet a bigger problem to be solved. That is, the heat generated from the deflection coil causes the thermal expansion of the wire member itself and the thermal expansion of the bobbin on which the deflection coil is attached. Furthermore, the heat generated from the deflection coil causes thermal expansion of ferrite pole pieces which form a projection lens. These thermal expansions change the deflecting direction and the deflection efficiency of the deflection coil and the lens magnetic field strength, and deteriorate the accuracy of the deflection position. The thermal expansion of the magnetic pole in particular introduces an error in the origin of the deflection coordinate and focal error.
Particularly in the case of an exposure which is made while a stage carrying the wafer continuously moves, an alignment mark on the wafer is detected prior to the exposure and the exposure is started after determining correction coefficients for the exposure. For this reason, the deterioration of the accuracy of the exposure position, the error in the origin of the deflection coordinate and the focal error which occur during the exposure are fatal to the quality of the exposure.
FIG. 3 shows measured results of deviation components of the beam position when the main deflector is caused to generate heat continuously. In FIG. 3, (A) is a graph showing the change of the amount of error of the offset position with time, (B) is a graph showing the change of the positional error in the rotation direction with time, and (C) is a graph showing the change of the positional error in the gain direction with time.
When the main deflector is caused to generated heat continuously, these amounts of errors will change up to large values as indicated by dotted lines in FIG. 3. As may be seen from FIG. 3, these amounts of errors are fatal to the exposure apparatus which exposes patterns on the submicron order.
Accordingly, it is conceivable to cool the electromagnetic deflection coils which form the main deflector, so as to prevent the drift by suppressing the heating. The present inventors actually made electromagnetic deflection coils having a large cooling capacity and studied the results.
The tested electromagnetic deflection coil employed a double-structure bobbin which is made up of an inner bobbin and an outer bobbin to support the electromagnetic deflection coil. The wire member of the electromagnetic deflection coil was wound in the radial direction in one layer and overlapped in the rotating direction in an arcuate manner to form a desired number of coil turns. The electromagnetic deflection coil was then bent along a cylindrical surface in the form of a saddle shape.
In addition, saddle shape coils having different radii of curvatures were made. The coil having the smaller radius of curvature was fixed on the outer circumference of the inner bobbin, and the coil having the larger radius of curvature was fixed on the inner circumference of the outer bobbin. A space was formed between the inner and outer coils so as to form a passage for flowing a coolant in the direction of the optical axis.
The bobbin was formed to a structure which is independent or integral to the plurality of stages of the coils, and was made of a material including quartz as the main component and having a small coefficient of linear expansion. Pure water or He gas was used as the coolant, and the coolant was forcibly circulated.
The electromagnetic deflection coils of the main deflector were cooled efficiently by the above arrangement, so as to reduce the thermal conduction to the parts such as the pole pieces. It was thought that the temperature rise of the structure will be extremely small by the efficient radiation.
The positional changes in the electron beam for the case where the above described electromagnetic deflection coils are used are indicated by solid lines in FIG. 3. In FIG. 3, a curve f1 in (A) shows the amount of error of the offset position as a function of time when cooling was made, a curve f2 in (B) shows the positional error in the rotating direction as a function of time when the cooling was made, and a curve f3 in (C) shows the positional error in the gain direction as a function of time when the cooling was made.
However, as may be seen from FIG. 3, even though the electromagnetic coils were cooled, the changes in the beam position were only reduced to one half of the case where no cooling was made.
For example, the amount of error of the offset position was approximately 0.5 .mu.m after approximately 3 minutes from the start of the exposure when no cooling was made. But even when the cooling was made, the amount of error in the offset position was only reduced to approximately 0.3 .mu.m. The amount of error of the offset position gradually saturated with time, and the amount of error was approximately 0.4 .mu.m after the saturation which was not within a tolerable range.
In addition, compared to the case where no cooling was made, the positional error in the rotating direction was only reduced to approximately one-half even when the cooling was made. The reduction of the positional error in the gain direction by making the cooling was even smaller compared to the case where no cooling was made.
In other words, even though the cooling efficiency improved with respect to the heat generation, it was evident that the cooling was incomplete and that the temperature change in the structure occurred due to the heat generation. It is conceivable to increase the cooling capacity of the coolant so that the amount of heat generated can be neglected, but there is a possibility of introducing mechanical vibration or the like due to the increased flow rate if the flow rate of the coolant is increased. On the other hand, there is a limit to increasing the flow rate of the coolant. It is also conceivable to further reduce the thermal conduction from the electromagnetic deflection coils to the pole pieces and the like, but there is of course a limit to doing this.
The temperature rise of the pole pieces with respect to the generation of heat by the electromagnetic deflection coils was also measured. It was found that the temperature rise is approximately 1.5.degree. C. in 10 minutes when no cooling was made and approximately 0.3.degree. C. in 10 minutes when the cooling was made. The temperature rise saturated in approximately 10 minutes, and the saturation values greatly differ between the case where no cooling is made and the case where the cooling is made. That is, the cooling effect can be seen.
However, the cooling effect is far from sufficient for the purpose of satisfactorily improving the accuracy of the exposure system. The temperature change in the initial stage in particular is not very large, and the cooling effect within the time of approximately 3 minutes from the start of the exposure is not very notable with respect to the drift of the beam position.
In order to obtain the high accuracy required by the semiconductor integrated circuits, the tolerable range of the temperature change of the structure due to the heat generated by the main deflector should be less than 0.1.degree. C. However, the cooling described above cannot realize such a small tolerable range.